Screening for immune-related biomarkers associated with myasthenia gravis and dilated cardiomyopathy based on bioinformatics analysis and machine learning

Abstract

Background: We aim to investigate genes associated with myasthenia gravis (MG), specifically those potentially implicated in the pathogenesis of dilated cardiomyopathy (DCM). Additionally, we seek to identify potential biomarkers for diagnosing myasthenia gravis co-occurring with DCM.

Methods: We obtained two expression profiling datasets related to DCM and MG from the Gene Expression Omnibus (GEO). Subsequently, we conducted differential gene expression analysis and weighted gene co-expression network analysis (WGCNA) on these datasets. The genes exhibiting differential expression common to both DCM and MG were employed for protein-protein interaction (PPI), Gene Ontology (GO) enrichment analysis, and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. Additionally, machine learning techniques were employed to identify potential biomarkers and develop a diagnostic nomogram for predicting MG-associated DCM. Subsequently, the machine learning results underwent validation using an external dataset. Finally, gene set enrichment analysis (GSEA) and machine algorithm analysis were conducted on pivotal model genes to further elucidate their potential mechanisms in MG-associated DCM.

Results: In our analysis of both DCM and MG datasets, we identified 2641 critical module genes and 11 differentially expressed genes shared between the two conditions. Enrichment analysis disclosed that these 11 genes primarily pertain to inflammation and immune regulation. Connectivity map (CMAP) analysis pinpointed SB-216763 as a potential drug for DCM treatment. The results from machine learning indicated the substantial diagnostic value of midline 1 interacting protein1 (MID1IP1) and PI3K-interacting protein 1 (PIK3IP1) in MG-associated DCM. These two hub genes were chosen as candidate biomarkers and employed to formulate a diagnostic nomogram with optimal diagnostic performance through machine learning. Simultaneously, single-gene GSEA results and immune cell infiltration analysis unveiled immune dysregulation in both DCM and MG, with MID1IP1 and PIK3IP1 showing significant associations with invasive immune cells.

Conclusion: We have elucidated the inflammatory and immune pathways associated with MG-related DCM and formulated a diagnostic nomogram for DCM utilizing MID1IP1/PIK3IP1. This contribution offers novel insights for prospective diagnostic approaches and therapeutic interventions in the context of MG coexisting with DCM.

Keywords: Diagnostic value; Dilated cardiomyopathy; Immune cell infiltration; Machine learning algorithms; Myasthenia gravis.

Graphical abstract

Fig. 1

Flow chart of this study design. DCM: Dilated cardiomyopathy, MG: Myasthenia gravis, WGCNA: Weighted gene co-expression network analysis, DEGs: Differentially expressed genes, CGs: Common genes, ROC: Receiver operating characteristic, DCA: Decision curve analysis.

Fig. 2

Identification of key module genes in the integrated DCM and MG dataset through WGCNA. A: Dendrogram of gene clustering in DCM, with different colors representing different modules. B: Correlation between module eigengenes and DCM, where blue indicates negative correlation and red indicates positive correlation. C: Dendrogram of gene clustering in MG, with different colors representing different modules. D: Correlation between module eigengenes and MG, where blue indicates a negative correlation and red indicates a positive correlation. DCM: Dilated cardiomyopathy, MG: Myasthenia gravis, WGCNA: Weighted gene co-expression network analysis. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 3

Differential expression analysis of the DCM and MG datasets. A: Heatmap of the top 25 upregulated and 25 downregulated DEGs in the DCM dataset. B: Volcano plot of DEGs in the DCM dataset, with green indicating downregulation and red indicating upregulation. C: Heatmap of the top 25 upregulated and 25 downregulated DEGs in the MG dataset. D: Volcano plot of DEGs in the MG dataset, with green indicating downregulation and red indicating upregulation. E: Venn diagram showing the intersection of DEGs in DCM and MG, named CGs. F: Key genes obtained by aggregating and de-duplicating WGCNA key module genes in DCM and MG, intersected with CGs using a Venn diagram, named Hub genes. DCM: Dilated cardiomyopathy, MG: Myasthenia gravis, DEGs: Differentially expressed genes, CGs: Common genes. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 4

GeneMANIA database analysis and functional enrichment analysis based on CGs. A: GeneMANIA database analysis of CGs, showing a network with a total of 31 genes and 721 connections, including Co-expression, Physical Interaction, Shared Protein Domains, and Predicted networks. B: Circular plot of the results of GO enrichment analysis based on CGs. C: Bubble plot of the results of KEGG enrichment analysis based on CGs. CGs: Common genes, GO Gene ontology, KEGG: Kyoto Encyclopedia of Genes and Genomes.

Fig. 5

Screening of potential small molecule compounds for DCM treatment using connectivity map analysis. A: Heatmap displaying the top 10 compounds with the highest enrichment scores across 9 cell lines based on connectivity map analysis. B: Descriptions of the top 10 compounds. C: Chemical structures of these 10 compounds.

Fig. 6

Screening of potential diagnostic biomarkers for MG-associated DCM using machine learning methods. A–B: Lasso regression analysis of the 6 hub genes to calculate the minimum value (A) and λ value (B) for diagnostic biomarkers. C–D: Random forest algorithm analysis of the 6 hub genes, with a Random forest plot generated; selection of biomarkers with Mean Decrease Gini scores greater than 6. E: Expression patterns of the 6 hub genes in the DCM dataset GSE57338. F: Expression patterns of the 6 hub genes in the MG dataset GSE85452. G: Venn diagram showing the upregulated genes expressed by the 6 Hub genes in the DCM and MG datasets, respectively. H: Venn diagram showing the downregulated genes expressed by the 6 hub genes in the DCM and MG datasets, respectively. DCM: Dilated cardiomyopathy, MG: Myasthenia gravis.

Fig. 7

Development and efficacy evaluation of the diagnostic nomogram model. A: Logic regression analysis of 3 genes, including C3AR1, PIK3IP1, and MID1IP1, followed by further screening to obtain the PIK3IP1 and MID1IP1 genes as the two key genes for constructing the diagnostic nomogram. B: Calibration curve of the nomogram model predictions in MG-associated DCM, where the dashed line labeled "Ideal" represents the standard curve, representing perfect predictions of the ideal model. The dotted line labeled "Apparent" represents the uncalibrated predicted curve, while the solid line labeled "Bias-corrected" represents the calibrated predicted curve. C: ROC ROC curve for the diagnostic performance of the two candidate biomarkers (PIK3IP1 and MID1IP1). D: DCA for the nomogram model. The black line is labeled as "None," representing the net benefit of the assumption that no patients have DCM. The grey line is labeled as "All," indicating the net benefit of the assumption that all patients have DCM, and the purple line is labeled as "Nomogram," representing the net benefit of the assumption that MG-related DCM cases are identified based on the diagnostic value of DCM predicted by the nomogram model. E–G: Calibration curve, ROC curve, and DCA decision curve for the nomogram in the external dataset GSE29819 of DCM. DCM: Dilated cardiomyopathy, MG: Myasthenia gravis, ROC: Receiver operating characteristic, DCA: Decision curve analysis. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 8

Single gene GSEA. A–F: GSEA results of PIK3IP1 in the DCM dataset GSE57338. G–L: GSEA results of MID1IP1 in the DCM dataset GSE57338. GSEA: Gene set enrichment analysis, DCM: Dilated cardiomyopathy.

Fig. 9

Analysis of immune cell infiltration in DCM and MG. A: Box plot comparing the infiltration of 22 immune cell types between the DCM group and the control group. B: Association between differentially infiltrated immune cells in DCM and the two hub genes, at a threshold of p < 0.05. C: Box plot comparing the infiltration of 22 immune cell types between the MG group and the control group. D: Association between differentially infiltrated immune cells in MG and the two hub genes, at a threshold of p < 0.05. DCM: Dilated cardiomyopathy, MG: Myasthenia gravis.